Image to Text: Extract Text From Any Picture, Free

Most online converters upload your picture to a server, run OCR there, and send text back. This one does not. The recognition model — a compact, modern OCR engine in the PaddleOCR family — is downloaded to your browser once and then runs locally in a background worker every time you convert an image.

That design has two practical consequences. First, privacy is structural, not a promise: because the conversion happens in your tab, there is no server that ever holds a copy of your image. Nothing about the file — not the pixels, not the filename — is transmitted. Second, it keeps working offline once the model is cached, so a captured receipt or screenshot converts even without a connection.

You do not have to take our word for it. Open your browser's developer tools, switch to the Network tab, and convert an image. You will see the model file download on the first run (and load from cache afterward), but you will not see your image being uploaded anywhere. The bytes stay on the device.

Screenshots are the single most common thing people run through this tool, and they are also what it reads best. A screen capture is crisp, evenly lit, and rendered at native pixel density, which is close to ideal input for OCR.

Typical screenshot to text jobs this handles cleanly:

• Copying an error message or stack trace that a program renders as an image or won't let you select
• Lifting text out of a chat, a social post, or a comment thread
• Grabbing code, a command, or a config snippet from a video or slide
• Pulling numbers off a dashboard, chart label, or spreadsheet capture
• Turning a screenshot of a PDF or paywalled page back into editable text

On macOS use Shift-Command-4, on Windows use the Snipping Tool (Windows-Shift-S), then drop the capture straight onto this page. Because the text in a screenshot is already sharp, accuracy on these is usually near-perfect.

Being honest about limits saves you time, so here is the real picture.

Reads excellently: printed and typeset text of almost any kind — books, receipts, invoices, forms, labels, signs, slide decks, and screenshots. Clear photos of a document page work well too, as long as the text is in focus and reasonably straight.

Struggles: handwriting. Cursive and casual handwritten notes are genuinely hard for in-browser OCR, and results will be poor — this tool is not the right one for them. We are building a dedicated handwriting-to-text tool that uses a different, heavier model for exactly this case; it is coming soon.

A few tips for better results: photograph documents straight-on rather than at an angle, get the whole text block in frame, and avoid glare and shadows. If a photo is blurry, retake it — sharper input beats any amount of post-processing. Very small or very low-contrast text may drop characters, so crop in close when you can.

Languages at launch: English, Spanish, and Indonesian. The engine recognizes the Latin-script characters, accents, and punctuation used across these languages, so Spanish tildes and accented vowels come through correctly. More languages are planned, and the tool will pick up the right model automatically as they are added.

Image formats: JPG/JPEG, PNG, WebP, and BMP are read directly. HEIC/HEIF — the format iPhones use by default — can't be decoded in the browser yet, so export or share those photos as JPG first (on iPhone, the Share sheet does this for you, or set Settings → Camera → Formats → Most Compatible to shoot JPG). For multi-page documents, a PDF is usually a better fit than an image; use our PDF to text tool for those.

There is no file-size cap imposed by a server, because there is no server in the loop — the practical limit is your device's memory. Everyday photos and screenshots are nowhere near that limit.